The choice to avoid the C: (System) drive for anything had a ton of merit back in the old days. If you have a small C: drive, avoid it, but if you have a large drive, it’s still useful for some things.
The main reason for avoiding C: was that, by default, you have a swapfile there (for virtual memory). That’s still there, but if you have a decent amount of memory, and don’t have too many apps going at once during even a complex render, it won’t get much use.
The other, not-as-often-understood reason is speculative loading. When you start a Windows application, the main bits of the application load, but other pieces, not necessarily. This can be DLLs that only load on demand, or it can parts of the main app that are marked in the process page tables as loaded but invalid… forcing the OS to load these pieces only when needed. This can be an issue for realtime work, but it’s not much of a problem for video rendering.
The core reason swapping might be an issue is simple… it increases disc seeking. An HDD drive that gives you 100MB/s performance from one file won’t give you 50MB/s each from two files, but something less. Each time it has to move heads from one file to another (eg, seeking), that’s time not spent transferring data.
However, this isn’t just a concern for the C: drive, but for any drive.. the more you load up on a single drive for a project, the more seeking you’ll experience, and the slower your effective drive speed will be. But really, as long as you see 100% CPU use while rendering, the drive’s probably still fast enough. When it starts to drop, that’s usually an indication that HDD I/O is now the bottleneck (certainly, other possible bottlenecks exist, like very CPU intensive plug-ins).
I keep a “scratch” drive along with C: and the project drive(s), which may contain small projects, becomes my default output drive for CineForm conversions, and gets used to offload assets from the main project if I have enough stuff in there to slow down a render.
RAID isn’t a bad idea, but it’s not always as fast as everyone thinks. And I dislike RAID0, as that’s less reliable than a single drive (eg, your drive fails based on the minimum time-to-failure of a set of drives). RAID is faster at reads, transfer-wise, but actually slower at seeking (since both drives have to seek, and you’re always going to have one slightly faster than the other, so you always wait for both). If Vegas is grabbing data in large enough chunks per file, the RAID will definitely help, since in theory, you need only half the number of logical seeks per file. This is the likely situations, since HDDs are only really fast if you grab data in large chunks, much larger than sector sized chunks.
For writes, RAID0 is always slower than the original drive, RAID1 slower still. This is, again, because you have to sync up both drives before the write can complete. Of course, if you have a “hardware” RAID (there’s a CPU on a PCI card somewhere running the RAID BIOS software, rather than using the NTFS stripeset or mirror set), this may well be buffered and hidden from you on writes. And of course, RAID5 is slower still, though again, with a good controller, this is hidden on writes.
I use plain old SATA drives: 1.5GB for C:, 1.5GB for D: (the “scratch” drive), and usually 500MB-1GB for the project drive. I also have an 8TB RAID on Firewire, but that’s used to hold older projects, DVD/BD sets, photos, that kind of thing. FW800 does not compare to SATA in performance… but it’s also practical to have that much storage [a] external and [b] not subject to the same power supply that powers your PC.
-Dave